Help with 'light leak'?

[sigurd]

Ummm... it is caused by "transistors" in a generic sense I suppose... Basically it is caused by heat given of from the chip's readout amplifier while integrating an image. What this means is that during the exposure the readout amplifier is left on: hence the heat/light. Some chips have this problem all the time, the old Canon 300D DIGIC chip for instance... However almost all astronomical cameras have that amplifier turned off during the exposure, and it is only on during readout. if it glows in one mode and not the other, my guess is that the driver doesn't power off the amp in the "non-fast" mode, which it probably ought to.

-esy

[pfile]

IDK, that amount of glow seems much less than the "sense amps always on" kind of glow. i don't know why fast readout mode would require the sense amps to be on during the image capture, while slow readout mode would not. and if it's a bug, it would seem to me it's one that would have been fixed pretty soon after the camera was released to customers. my money's still on a higher clock rate during column/row shift causing that corner of the chip to heat up a little bit more that it does during slow readout.

here's a tidbit from the SBIG ST-8300 manual:

The readout (or digitization) of images from CCDs is a sequential process whereby every pixel is digitized one-after-another until the whole image has been digitized. The readout starts at pixel (1,1) in the top-left corner and proceeds to pixel (3326, 2504) in the lower-right hand corner. As each pixel in a row is readout the pixels to the right of it within the same row are shifted to the left one position in preparation for the next pixel readout. The next pixel is then readout and the process repeated until every pixel in the first row has been digitized.
At this point the whole CCD is shifted up one row and digitization starts with the left-most pixel of the second row. The readout glow on the left of the images is a buildup of light from a glow from the preamplifier structures in the upper-left hand corner of the CCD while rows are queuing up for readout.

here's a good page on the basics of CCD readout:

http://spiff.rit.edu/classes/phys445/lectures/ccd1/ccd1.html

[sigurd]

We could probably go around on this forever, but it's not an issue with the chip itself, but rather the way in which the camera manufacturer is using it.

My FLI ML8300 (based on the same Kodak chip), absolutely does not have  this amplifier glow. Readouts are performed at 8 MHz at all times with this camera (that's a full frame in approx. 2 seconds).

What SBIG is stating in their manual is factually correct, but misleading. It leaves the impression that the the amount of glow you are seeing is caused by the amplifier during readout. I don't see how this could be the case. If it were, the slower readout mode would have to show more glow (since the glow amount would be dependent on time, though possibly their "turbo mode" would run the amplifier "hotter" given its faster speed). It's seems possible that the readout could case some glow, but I would think the amount of glow would be reversed. Of course all this means that the glow you are having should be removable with a good bias set (or a good set of darks, if you don't use a scalable dark reduction strategy).

I'm usually familiar with amp glow in this context:

"Due to the faintness of astronomical targets in general long exposures are often required in order to collect enough light. During these long integrations noise can also be added to the image being built up in the CCD. Thermal noise effects pixels unevenly with some pixels being termed ‘hot’ and saturate quickly while others more slowly accurate dark current. These effects are reduced by cooling the CCD. The amplifier on the CCD chip produces electroluminescence that appears as a glow coming from one corner of the CCD.   Switching off the amplifier until it is needed for read out can eliminate this."
-http://www.google.com/url?sa=t&source=web&cd=2&ved=0CBsQFjAB&url=http%3A%2F%2Fwww.pmdo.com%2Fartdownloads%2FIntroduction.pdf&rct=j&q=ccd%20amp%20glow%20from%20readout%3F&ei=fnlMTdXqJoSClAfm9sAu&usg=AFQjCNGtE4wGJhe7p0K6_hH-c4BqUnlcpw&sig2=wDSNIdDL54qoMhvd7YwMgg

and also here:

http://www.qucam.com/emccd/L3Poster.pdf

To me this all really depends on the design of the camera. the ST-8300 clearly exhibits it in one of your readout modes. My FLI does not.

-esy

[pfile]

We could probably go around on this forever, but it's not an issue with the chip itself, but rather the way in which the camera manufacturer is using it.

agreed, but that's what the internet is for 

My FLI ML8300 (based on the same Kodak chip), absolutely does not have  this amplifier glow. Readouts are performed at 8 MHz at all times with this camera (that's a full frame in approx. 2 seconds).

What SBIG is stating in their manual is factually correct, but misleading. It leaves the impression that the the amount of glow you are seeing is caused by the amplifier during readout. I don't see how this could be the case. If it were, the slower readout mode would have to show more glow (since the glow amount would be dependent on time, though possibly their "turbo mode" would run the amplifier "hotter" given its faster speed). It's seems possible that the readout could case some glow, but I would think the amount of glow would be reversed. Of course all this means that the glow you are having should be removable with a good bias set (or a good set of darks, if you don't use a scalable dark reduction strategy).

i dont know what the clock rates are for the different readout speeds, but power consumption (and thus heat) in digital circuits goes linearly with clock rate. i had the same thought about the glow being reversed, since as the lines are read out you'd think it would be hotter toward the end of the shift and not toward the beginning. thus as the pixels farthest from the amplifier get closer to the amp, they are artificially brightened by the heat in that corner of the chip. the glow would appear opposite the physical location of the amplifier.

at any rate, given that you don't see this with your camera and it has the same chip, i wonder if this means the OP's camera has either a design or a one-off defect where the peltier cooler is somehow not well-bonded to the CCD, letting enough heat build up during readout such that the glow is visible.

also as you point out this glow is very removable with calibration frames. i don't think there's any reason to fear it, given that it is not uncommon for CCDs and in fact pretty minimal compared to the example of the 300D. lots of people do good work with the 300D despite the amp glow.

[sigurd]

Well, you know, I think your first point is basically the answer... if the glow were caused by the readout itself it would appear in the columns opposite the amplifier. I think that pretty conclusively shows that they must just leave the amp on during the entire integration when using "Turbo Mode". This means it must be removable with darks (rather than bias). My guess is "Turbo Mode" is provided for focusing, and hence one shouldn't get too worked up about the glow, since notionally those kinds of exposures would rarely exceed 1s. I don't think it's a defect in the camera, it seems like a design choice. I would just stick with the "slow" exposure/readout mode for long integrations.

I'm not sure what the clock rate of the SBIG is either, but I do know it takes about 9 seconds to read out a full frame. so my guess is it's probably 4 MHz.

-esy